Error generating title

New Consortium Moves to Standardize Semantic Modeling for AI

A group of major data and AI vendors has announced the Open Semantic Interchange (OSI), an effort to create an open source standard for semantic data modeling. Participants include Snowflake, Salesforce, Alation, Atlan, Cube, DBT Labs, Mistral AI, Sigma and ThoughtSpot.

The goal: make data definitions consistent across tools so teams can integrate, discover and use data without the usual friction. For leaders, this translates to faster AI delivery, fewer disputes over metrics and higher trust in outcomes.

Why managers should care

Inconsistent data definitions slow projects, inflate costs and erode confidence in AI outputs. A shared semantic standard cuts the time your teams spend reconciling metrics and hunting for the "right" version of truth.

It also reduces vendor lock-in. If semantics are portable, you can switch tools or combine platforms without breaking business logic.

What a semantic model is (and why it's been a gap)

A semantic model defines business concepts and metrics in a consistent way-so "customer," "revenue," or "churn" mean the same thing wherever they appear. Many organizations don't have this layer, and many platforms ship with proprietary semantics that don't interoperate. The result is fragmented data, duplicated work and brittle analytics.

What's new with OSI

OSI's purpose is to standardize how semantics are defined, exchanged and reused across tools. Snowflake signaled its intent with the launch of Semantic Views in June and is co-leading the initiative with partners that heard the same customer pain: fragmented definitions blocking AI scale.

Expert perspective

Stephen Catanzano of Enterprise Strategy Group called a shared semantic standard "highly important," pointing to the way fragmented semantics create roadblocks for human and AI analysis. Kevin Petrie of BARC U.S. added that a unified semantic layer helps mitigate data quality issues by enabling AI to consume diverse inputs more reliably.

GenAI and agents raise the stakes

Generative AI and emerging autonomous agents need large volumes of consistent, high-quality data. Without shared definitions, systems produce inconsistent or untrustworthy results. Semantic modeling makes relevant data easier to find and combine; an open standard makes it portable across platforms and teams.

OSI goals

• Improve interoperability across tools and platforms through a shared semantic standard to make integrating and preparing data easier.
• Accelerate development and deployment of AI and analytics applications by standardizing how semantics are defined and exchanged.
• Streamline operations by reducing the need to reconcile conflicting semantic definitions and duplicate work across platforms.

Adoption risks and dependencies

Success depends on broad industry support. Analysts note that hyperscalers (AWS, Google, Microsoft) and cloud data platforms like Databricks need to participate-or OSI risks becoming another silo. On-premises matters too: research indicates roughly one-third of AI workloads still run on-prem.

What leaders should do now

• Appoint an executive owner for semantics (often within Data Governance or a CDO office).
• Inventory your critical metrics and definitions; document the conflicts across BI, data science and product teams.
• Pilot a semantic model in one business domain (e.g., revenue) and measure time saved in reconciliation.
• Ask current vendors for their OSI roadmap and timelines for import/export of OSI models.
• Set policy for semantic versioning, testing and approval before changes go live.
• Fund data modeling as a product, not a project-ongoing ownership, SLAs and release cadence.

Questions to press your vendors on

• Will you support the OSI specification for read/write, not just read-only?
• How will you map proprietary semantic layers to OSI without losing logic or lineage?
• What governance, validation and CI/CD exist for semantic changes?
• Can we track lineage from source data to business metric across platforms?
• How will your tools interoperate with agent frameworks and context protocols?

Metrics to track

• Time to reconcile metric disputes (target: weeks to days).
• Percentage of analytics and AI use cases using approved semantics.
• Number of duplicate metrics removed and definitions consolidated.
• Agent or model accuracy tied to semantic coverage (before/after).
• Cost and time impact of tool changes (with/without OSI-compliant semantics).

Timeline and what to expect

A working group has formed to deliver the open standard "quickly." Additional partners are being recruited. The first releases will likely focus on a vendor-neutral model specification, import/export formats and baseline governance patterns.

Related standards to watch

Open standards are also emerging for agents. For example, the Model Context Protocol defines how agents interact with data and tools. See the official site for details: Model Context Protocol. As agents spread, expect more open specifications for data integration and model evaluation.

Bottom line for management

Semantic consistency is now a strategic control point. OSI is a credible move to reduce friction, improve trust and speed up AI delivery. Push your teams and vendors to align early so you avoid rework later-and measure the impact in fewer disputes, faster releases and better model performance.

Upskill your teams

If you're building internal capability around AI, governance and data strategy, explore role-based learning paths here: Complete AI Training - Courses by Job.